INFORMATION SOCIETY IS 2017 - IJSkt.ijs.si/MarkoBohanec/pub/2017_IS_HeartMan-DSS.pdf · Slovenska...
Transcript of INFORMATION SOCIETY IS 2017 - IJSkt.ijs.si/MarkoBohanec/pub/2017_IS_HeartMan-DSS.pdf · Slovenska...
Zbornik 20. mednarodne multikonference
INFORMACIJSKA DRUŽBA – IS 2017 Zvezek A
Proceedings of the 20th International Multiconference
INFORMATION SOCIETY – IS 2017 Volume A
Slovenska konferenca o umetni inteligenci Slovenian Conference on Artificial Intelligence
Uredili / Edited by
Mitja Luštrek, Rok Piltaver, Matjaž Gams
http://is.ijs.si
9. - 13. oktober 2017 / 9th – 13th October 2017Ljubljana, Slovenia
Designing a Personal Decision Support System for
Congestive Heart Failure Management Marko Bohanec, Erik Dovgan, Pavel Maslov, Aljoša Vodopija,
Mitja Luštrek Jožef Stefan Institute
Department of Intelligent Systems, Department of Knowledge
Technologies Jamova cesta 39, 1000 Ljubljana,
Slovenia
{marko.bohanec, erik.dovgan, pavel.maslov, aljosa.vodopija,
mitja.lustrek}@ijs.si
Paolo Emilio Puddu, Michele Schiariti,
Maria Costanza Ciancarelli Sapienza University of Rome
Department of Cardiovascular, Respiratory, Nephrologic and Geriatric
Sciences Piazzale Aldo Moro 5, Roma 00185,
Italy
{paoloemilio.puddu. michele.schiariti}@uniroma1.it
Anneleen Baert, Sofie Pardaens,
Els Clays Ghent University
Department of Public Health De Pintelaan 185 – 4K3, 9000 Gent,
Belgium
{anneleen.baert, sofie.pardaens,
els.clays}@ugent.be
ABSTRACT
In this paper, we describe the design of the HeartMan Decision
Support System (DSS). The DSS is aimed at helping patients
suffering from congestive heart failure to better manage their
disease. The support includes regular measurements of patients’
physical and psychological state using a wristband and mobile
device, and providing advice about physical exercise, nutrition,
medication therapy, and environment management. In the paper,
an overall architecture of the DSS is presented, followed by a
more detailed description of the module for physical exercise
management.
Categories and Subject Descriptors
H.4.2 [Types of Systems]: Decision support.
J.3 [Life and Medical Sciences]: Medical information systems.
General Terms
Algorithms, Management, Measurement, Design, Human Factors
Keywords
Decision Support System, Personal Health System, Congestive
Heart Failure, Physical Exercise, Decision Models
1. INTRODUCTION Congestive heart failure (CHF) occurs when the heart is unable to
pump sufficiently to maintain blood flow to meet the body's needs
[1]. Symptoms include shortness of breath, excessive tiredness,
and leg swelling. CHF is a common, chronic, costly, and
potentially fatal condition [2]. In 2015 it affected about 40 million
people globally. In developed countries, around 2% of adults have
heart failure, increasing to 6–10% in age over 65.
HeartMan (http://heartman-project.eu/) is a research project
funded by the European Union’s Horizon 2020 research and
innovation programme. The project aims to develop a personal
health system to help CHF patients manage their disease. CHF
patients have to take various medications, monitor their weight,
exercise appropriately, watch what they eat and drink, and make
other changes to their lifestyle. The HeartMan system will provide
accurate advice on disease management adapted to each patient in
a friendly and supportive fashion. The DSS follows the best
medical practices [3] and is designed so that it never suggests
anything that would harm the patient.
In this paper, we present the design of the HeartMan Decision
Support System (DSS), which was finalised in June 2017 [3]. In
section 2, we describe the overall functionality and architecture of
the system, and define the roles of its modules that address (1)
physiological measurements, (2) physical exercise, (3) nutrition,
(4) medication, (5) environment management, and (6)
management of calendars and plans. In section 3, we focus on the
physical exercise module and present its most important
components for (1) patients’ physical capacity assessment, (2)
weekly exercise planning, and (3) daily exercise management.
2. DESIGN OF THE HEARTMAN DSS The HeartMan DSS aims at providing medical advice to CHF
patients using predictive models, clinical care guidelines and
expert knowledge. The purpose of a typical DSS is to passively
present information to decision makers so that they can make
maximally informed decisions. This DSS, however, is intended
for patients who have limited medical knowledge and are
consequently expected to follow guidelines with little discretion.
Because of that, the DSS actively provides advice to patients,
although it does offer choice where appropriate. In this way, it
belongs to the category of cooperative DSS [4].
Phone sensors
Health devices
Monitoring physical and psychological
state
Mobile application
Decision Support Modules
1. Physiological measurements
2. Physical exercise3. Nutrition4. Medication5. Environment
management6. Calendars and plans
Data management
Web interface
Figure 1: Overall architecture of the HeartMan DSS.
67
The overall architecture of the HeartMan DSS is shown in Figure
1. The system will use wrist-band sensors to monitor patient’s
physical activity, heart rate and some other physiological signs. In
addition, it will receive data from additional devices, such as
scales, smartphone and from the patient via the user interface of
the mobile application. This will allow the system to identify the
patient's current physical and psychological characteristics. This
data will be combined with patient’s health data to help them
decide on disease control measures. The advice will be tailored to
the patient's medical condition by adapting it to the patient's
psychological profile (such as normal, poorly motivated,
depressed, and anxious) and current health state. The advice will
be shown at the mobile app. A web-based interface will be
provided to the physician, too, who will be able to monitor the
patient’s health state and progress, and define or approve
parameters that affect the advice given to the patient.
The core of the HeartMan DSS are modules that interpret
patient’s data and make recommendations. There are six main
modules, which address the following aspects of health
management:
1. Physiological measurements: CHF patients should perform
various physiological measurements on a regular basis, such
as measuring their weight, blood pressure, heart rate, etc. This
raises the patients’ awareness of their health, provides
valuable information to their physicians, and provides inputs
to the DSS. For this purpose, the DSS reminds the patient to
regularly perform these measurements, provides the
functionality to carry them out, and manage the collected data.
2. Exercise: Physical conditioning by exercise training reduces
mortality and hospitalization, and improves exercise tolerance
and health-related quality of life. For this purpose, the DSS
provides a comprehensive exercise programme, which is
detailed later in section 3.
3. Nutrition: CHF patients should maintain their body weight
and take care of their diet, for instance, not eating too much
salt or drinking too much fluid. The DSS assesses the
patients’ nutrition behavior, educates them through a quiz and
provides advice towards a healthy diet.
4. Medication: Good adherence to medication therapy decreases
mortality and morbidity, and improves well-being of CHF
patients. For this purpose, the DSS reminds the patient to take
medications and assesses the patient’s adherence to the
medication scheme. For each medication, the patient may
obtain an explanation why the adherence is important.
5. Environment management: Environmental conditions, such as
temperature and humidity, may affect the patient’s feeling of
health. Combining both, the patient’s and environmental
conditions, the DSS advises the patient how to change the
environment to improve their health feeling.
6. Calendars and plans: Given all the DSS aspects
(measurement, exercise, nutrition, medication and
environment management) and many interactions between
them, it is important to sensibly arrange all the activities and
notifications, for instance not sending nutrition advice during
exercise or suggesting physical exercise after taking diuretics.
This DSS module thus coordinates the activities and arranges
all the plans into one single calendar.
To date, all these modules have been designed in terms of their
functionality, requirements, input data, processing, outputs, and
distribution between the client (patient’s mobile app) and the
server (the DSS in “the cloud”) [3].
3. PHYSICAL EXERCISE MODULE The HeartMan DSS administers a comprehensive exercise
programme. At the beginning, the DSS collects medical
information and assesses patient’s physical capacity in order to
plan the difficulty level of the exercises. Then, the DSS provides a
weekly set of endurance and resistance exercises, which increase
in difficulty as the patient becomes fitter. The DSS also guides the
patient during each exercise session: it checks whether the patient
is ready to start, then provides instructions, and finally asks the
patient to evaluate the exercise. The exercise module follows the
guidelines provided in [5] with minor modifications to fit in a
mobile application.
3.1 Physical Capacity Assessment Prior to starting using the HeartMan DSS, the patients should
perform a cardiopulmonary exercise (cycloergometry) test to
assess their physical capacity. Alternatively, when using the
system in a supervised, standardized setting, patients can perform
a 6-minute walking test. On this basis, the physical capacity of
each patient is assessed as “low” (less than 1 W/kg measured by
cycloergometry or less than 300 m walked in 6 minutes) or
“normal” (otherwise).
3.2 Weekly Exercise Planning The DSS system provides the patient with a combined endurance
and resistance exercise programme. Both types follow the same
principle described with four parameters: frequency (times per
week), intensity, duration and type. These parameters are
combined with the physical capacity to make a weekly exercise
plan for each patient. For instance, low-capacity patients start with
very light 10-15-minute endurance exercises twice per week.
According to the patient’s progress, these parameters may change
with time. In the HeartMan DSS, the progress is prescribed by
two models:
EnduranceFrequency: a model for suggesting weekly
frequency of endurance exercises;
EnduranceTime: a model for suggesting weekly time
boundaries of endurance exercises.
Both models are formulated using a qualitative multi-criteria
decision analysis method DEX [6]. Here, we illustrate the
approach describing the EnduranceFrequency model, whose
structure is shown in Figure 2.
DEXi EnduranceFrequency.dxi 15.5.17 Page 1
Scales Attribute Scale EnduranceFrequency 2x; 3x; 4x; 5x
Normative 2x; 3x; 4x; 5xCategory low; normalWeek 1; 2; 3; 4; 5; 6; 7; 8; 9; 10; 11; 12; 13; 14; 15; 16; 17; 18; 19; 20; 21; 22; 23; 24; more
Current 2x; 3x; 4x; 5xTransition decrease; stay; increase; automatic
MedicalAssessment decrease; stay; increase; automaticPatientsAssessment decrease; stay; increase; automatic
Figure 2: Structure of the EnduranceFrequency model.
The EnduranceFrequency model is aimed at suggesting the
frequency of exercises for the next week, based on the patient’s
physical capacity, week in the programme, current frequency, and
the possible physician’s and patient’s suggestions for the change.
In other words, the model takes into account both the normative
68
(as proposed by a general programme) and actual (as practiced by
the patient) frequency, leveraging the patient’s and physician’s
opinion about the suggestion for the subsequent week.
The overall recommendation, which is 2, 3, 4, or 5 times per
week, is represented by the root attribute EnduranceFrequency
(Figure 2). The recommendation depends on three sub-criteria:
1. Normative: Frequency as suggested according to the default
programme. It depends on the patient’s physical capacity
(“low” or “normal” Category) and the current Week. The
progression is defined by rules presented in Table 1.
2. Current: The frequency of exercises currently carried out by
the patient; it can run ahead or behind the Normative plan. In
order to make only small and gradual changes to the
frequency, Current is compared to Normative and only a one-
step change is suggested in each week.
3. Transition is an attribute that captures the patient’s wish and
the physician’s opinion about changing the frequency. The
possible values are “decrease”, “same”, “increase” or
“automatic”; the latter is meant to suggest the frequency
according to the normal plan, for instance, when neither the
patient or physician have given any suggestion. The patient’s
and physician’s suggestions are combined according to
decision rules shown in Table 2. The first two rules say that
whenever the patient or the physician suggest to decrease the
frequency, it should indeed be decreased (the symbol ‘*’
represents any possible value). Rules 3 and 4 suggest to keep
the current frequency whenever one of the participants
suggests so, unless the other participant suggests “decrease”
Rules 5 and 6 define a similar reasoning for “increase”. If
both participants have no particular suggestions, the
“automatic” transition according to the normal plan takes
place.
Table 1: Decision table defining the Normative frequency.
Category Week Normative
1 low <=4 2x
2 low 5–12 3x
3 normal <=6 3x
4 low 13–18 4x
5 normal 7–12 4x
6 low >=19 5x
7 normal >=13 5x
Table 2: Decision rules for Transition.
MedicalAssessment PatientsAssessment Transition
1 decrease * decrease
2 * decrease decrease
3 stay not decrease stay
4 not decrease stay stay
5 increase increase or automatic increase
6 increase or automatic increase increase
7 automatic automatic automatic
3.3 Daily Exercise Management Once a weekly plan has been established, the HeartMan DSS
assists the patient in carrying out their daily exercises. This
consists of four activities: (1) reminding the patient, (2) pre-
exercise checking, (3) exercise monitoring, and (4) post-exercise
assessment.
3.3.1 Reminding the patient Patients can choose the days when they want to exercise (e.g.,
every Tuesday, Thursday and Sunday). On these particular days,
the patients are at some predefined morning time reminded about
the daily exercise. Another reminder is issued if the exercise has
not been completed before the given afternoon time.
3.3.2 Before the exercise Before the start of each exercise session, the HeartMan DSS
checks if all prior-exercise requirements are met, and advises the
patients about safety. Figure 3 shows the decision model.
Figure 3: Pre-exercise assessment.
1. Information requirements: The blood pressure should have
been measured during the day. If not, the patients are
instructed to measure it. The pre-exercise heart rate is
measured automatically by the wristband; the system makes
sure that it is actually worn.
2. Medication requirements: Patients are asked to fill a check-list
of frequently seen side effects based on their medication
schemes and symptoms (e.g., dizziness and chest pain). On
this basis the DSS checks for any possible restrictions due to
medications or symptoms and suggests rescheduling the
session if necessary. The physician or nurse are contacted if
severe side effects are present. In the case of dizziness or chest
pain, patients are instructed to rest until the symptoms are no
longer present.
3. Physiological requirements: If all the requirements are met,
patients can start with the exercise, otherwise they are
instructed to repeat the measurements after five minutes of
rest. If after re-checking the measurements are still not within
safe limits, exercise is not allowed and patients are advised to
contact their physician or heart failure nurse.
Again, a DEX [3, 6] model is employed for assembling and
checking the medical conditions, which include medical intake,
comorbidities and current physical condition of the patient. The
structure and scales of the PreExerciseRequirements model are
shown in Figure 4. All attributes are binary (“yes”/“no” or
“not_met”/“met”). The values of the input attributes are
69
determined from patient data whenever the pre-exercise
requirements are checked (normally once per day before making
exercises). The subtrees of the model comprise four main groups
of possible reasons against participating in the exercises:
Blood coagulation: Whenever the patient takes anticoagulants
and there are symptoms indicating a possible bleeding: rash,
hemorrhages, or neurological symptoms.
Medication intake: Whenever one of the following
medications has been taken 2 hours or less before the
exercise: beta blockers, ACE inhibitors, ARBs, diuretics, or
loop diuretics.
Heart rate: Whenever the patient takes Digitalis and his/her
HR is less than 45 bpm.
Blood pressure: Whenever there are risks of hypertension
(taking ACE inhibitors or ARBs, and the patient has persistent
low blood pressure or persistent cough) or problems regarding
the systolic blood pressure (when the patient’s systolic blood
pressure is less than 105 and he/she recently took loop
diuretics). DEXi PreExerciseRequirements.dxi 16.5.17 Page 1
Scales
Attribute Scale PreExerciseRequirements not_met; met
BloodCoagulationReasons yes; noTakesAnticoagulats yes; noPossibleBleeding yes; no
Rash yes; noHemorrhages yes; noNeurologicalSymptoms yes; no
MedicationIntakeReasons yes; noIntake<2hours yes; noExercisePreventionMedications yes; no
TakesBetaBlockers yes; noTakesACEInhibitors yes; noTakesARBs yes; noTakesDiuretics yes; noTakesLoopDiuretics yes; no
HeartRateReasons yes; noTakesDigitalis yes; noHR<45 yes; no
BloodPressureReasons yes; noHypertensionReasons yes; noTakesACEInhibitors yes; noTakesARBs yes; noPersistentLowBloodPressure yes; noPersistentCough yes; no
SystolicPressureReasons yes; noTakesLoopDiuretics yes; noTookLoopDiuretics yes; noSYS<105 yes; no
Figure 4: Structure of the PreExerciseRequirements model.
3.3.3 During the exercise If the exercise is allowed, a list of exercises is shown to the
patient, who can then select the preferred exercise. After selecting
the exercise, a detailed description (text or graphical) regarding
the exercise is provided.
During the exercise, the heart rate and systolic blood pressure are
continuously measured by the wristband. The patients are advised
to stop the exercise in case of symptoms or measurements lying
outside of prescribed safety margins. If the heart rate is within the
safety limits, but too low or too high, the patent is advised to
increase or decrease the intensity, respectively. The system also
advises the patients about the exercise duration and is capable of
recognizing a premature ending.
3.3.4 After the exercise After completing the exercise, the patients can rate their feeling of
intensity (very light, light, moderate, intense, very intense). Then
the system assesses the exercise based on measurements recorded
during the exercise. It checks if the exercise was prematurely
finished and if the intensity was on average in the prescribed
limits. The system takes into account this information when
assessing the adherence to the exercise plan and the patient’s
improvement. Independent of this, the exercise is shown as
completed and the weekly plan is updated.
4. CONCLUSIONThis paper described the design of the HeartMan DSS that is
concerned with “medical” interventions (i.e., interventions that try
to improve the patients’ physical condition as opposed to
psychological). The DSS is based on clinical guidelines for the
self-management of CHF, additional medical literature and expert
knowledge from the project consortium. The DSS is designed in
terms of process models (the order of actions and questions) and
decision models (how to make some complex decisions –
branching in the process models) for five main topics of CHF
management: physiological measurements, exercise, nutrition,
medication, and environment management.
The DSS is currently being integrated in the overall HeartMan
platform. A comprehensive validation involving 120 CHF patients
(of whom 40 are controls without the DSS) is planned for 2018.
5. ACKNOWLEDGMENTSThe HeartMan project has received funding from the European
Union’s Horizon 2020 research and innovation programme under
grant agreement No 689660. Project partners are Jožef Stefan
Institute, Sapienza University, Ghent University, National
Research Council, ATOS Spain SA, SenLab, KU Leuven, MEGA
Electronics Ltd and European Heart Network.
6. REFERENCES[1] Heart failure. Health Information. Mayo Clinic. 23
December 2009. DS00061. http://www.mayoclinic.org/
diseases-conditions/heart-failure/basics/definition/con-
20029801.
[2] McMurray, J.J., Pfeffer. M.A. 2005. Heart failure. Lancet
365 (9474): 1877–89. DOI=
https://doi.org/10.1016%2FS0140-6736%2805%2966621-4.
[3] Bohanec, M., Vodopija, A., Dovgan, W., Maslov, P.,
Luštrek, M., Puddu, P.E., Clays, E., Pardaens, S., Baert, A.
(2017). Medical DSS for CHF. HeartMan Deliverable D4.1.
[4] Turban, E., Sharda, R., Delen, D., King, D. (2010). Business
Intelligence. 2nd Edition, Prentice Hall.
[5] Piepoli, M.F., Conraads, V., Corrà, U., et al. (2011):
Exercise training in heart failure: from theory to practice. A
consensus document of the Heart Failure Association and the
European Association for Cardiovascular Prevention and
Rehabilitation. European Journal of Heart Failure 13, 347–
357.
[6] Bohanec, M., Rajkovič, V., Bratko, I., Zupan, B., Žnidaršič,
M. (2013): DEX methodology: Three decades of qualitative
multi-attribute modelling. Informatica 37, 49–5
70